DeepLab2 is a TensorFlow library for deep labeling, aiming to provide a unified and state-of-the-art TensorFlow codebase for dense pixel labeling tasks.
I am trying to export a kMaX model to onnx format or a trt engine.
The error accures when calling trt.TrtGraphConverterV2
Both methods fail with the same error:
StatefulPartitionedCall/kMaXDeepLab/kmax_pixel_decoder/kmax_pixeldecoder_0/block1/attention/height_axis/query_rpe/Gather/axis is not unique
I am using kmax_meta_r50_os32.textproto with a dataset similar to cityscape.
Hi,
I am trying to export a kMaX model to onnx format or a trt engine. The error accures when calling
trt.TrtGraphConverterV2Both methods fail with the same error:StatefulPartitionedCall/kMaXDeepLab/kmax_pixel_decoder/kmax_pixeldecoder_0/block1/attention/height_axis/query_rpe/Gather/axis is not uniqueI am using
kmax_meta_r50_os32.textprotowith a dataset similar to cityscape.The same error was mentioned in https://github.com/google-research/deeplab2/issues/50 and in https://github.com/google-research/deeplab2/issues/82
I have read the answers but for me it is stillnot clear how to fix the issue.
I only need support a batch size of 1.